Metraipolis › archive › dc_5a29be154b82

Tsubame: Tree Replay for Diffusion-Based Speculative Decoding — arXiv:2609.33652

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33652
License (as archived)
CC0-1.0
Archived
2026-10-07T21:01:38.863Z
SHA-256 of the archived text
e420def57829c154e2d1251cd4b4993db39b4392f69cf22aa9b734d97d5128a6
Stable URL of this page
https://metraipolis.com/documento/dc_5a29be154b82

How to cite

Tsubame: Tree Replay for Diffusion-Based Speculative Decoding — arXiv:2609.33652 — arXiv (en), https://arxiv.org/abs/2609.33652 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5a29be154b82 (Metraipolis archive dc_5a29be154b82, sha256 e420def57829c154e2d1251cd4b4993db39b4392f69cf22aa9b734d97d5128a6)
BibTeX
@misc{dc5a29be154b82,
  title = {Tsubame: Tree Replay for Diffusion-Based Speculative Decoding — arXiv:2609.33652},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33652},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5a29be154b82; sha256 e420def57829c154e2d1251cd4b4993db39b4392f69cf22aa9b734d97d5128a6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33652. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Context-aware dynamic trees allocate the speculative decoding budget according to draft path probabilities, adapting their depth and branching to the current context. Under stochastic decoding, however, we find that this structural advantage does not always compensate for the acceptance gains of random sampling paired with advanced verification, and such dynamic trees can fall behind sampled chains in some settings. These trees grow their topology from the candidates themselves, so the tokens submitted for verification are typically the deterministic high-score tokens selected during construction. This coupling is not inherent: once the topology is fixed, its nodes can be repopulated by sampling, allowing dynamic trees to retain their structural advantage while also benefiting from random sampling and advanced verification. Diffusion-based drafters make this practical, as their parallel outputs or lightweight conditional corrections allow candidates to be regenerated cheaply after the complete topology is known. We introduce Tsubame, a two-pass tree speculative decoding framework for diffusion-based drafters. The first pass plans and freezes a context-aware topology using draft path scores; the second replays the fixed topology, sampling the tokens that populate its nodes to form the candidate tree for verification. We prove that Tsubame is lossless under compatible sampling and verification strategies. Experiments across three diffusion-based drafters, six datasets, and multiple candidate budgets show that Tsubame improves acceptance length and throughput over deterministic trees, including settings where it reverses their disadvantage against sampled chains.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.33652 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:02.716Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).